Articles

Estimates and 25-year trends of the global burden of disease

attributable to ambient air pollution: an analysis of data
from the Global Burden of Diseases Study 2015
Aaron J Cohen*, Michael Brauer*, Richard Burnett, H Ross Anderson, Joseph Frostad, Kara Estep, Kalpana Balakrishnan, Bert Brunekreef,
Lalit Dandona, Rakhi Dandona, Valery Feigin, Greg Freedman, Bryan Hubbell, Amelia Jobling, Haidong Kan, Luke Knibbs, Yang Liu, Randall Martin,
Lidia Morawska, C Arden Pope III, Hwashin Shin, Kurt Straif, Gavin Shaddick, Matthew Thomas, Rita van Dingenen, Aaron van Donkelaar,
Theo Vos, Christopher J L Murray, Mohammad H Forouzanfar†

Summary
Background Exposure to ambient air pollution increases morbidity and mortality, and is a leading contributor to Lancet 2017; 389: 1907–18
global disease burden. We explored spatial and temporal trends in mortality and burden of disease attributable to Published Online
ambient air pollution from 1990 to 2015 at global, regional, and country levels. April 10, 2017
http://dx.doi.org/10.1016/
S0140-6736(17)30505-6
Methods We estimated global population-weighted mean concentrations of particle mass with aerodynamic
This online publication has been
diameter less than 2·5 µm (PM2·5) and ozone at an approximate 11 km × 11 km resolution with satellite-based corrected. The corrected version
estimates, chemical transport models, and ground-level measurements. Using integrated exposure–response first appeared at thelancet.com
functions for each cause of death, we estimated the relative risk of mortality from ischaemic heart disease, on April 19, 2018
cerebrovascular disease, chronic obstructive pulmonary disease, lung cancer, and lower respiratory infections See Comment page 1862
from epidemiological studies using non-linear exposure–response functions spanning the global range *Joint first authors
of exposure. †Senior author
Health Effects Institute,
Findings Ambient PM2·5 was the fifth-ranking mortality risk factor in 2015. Exposure to PM2·5 caused 4·2 million Boston, MA, USA
(A J Cohen DSc); University of
(95% uncertainty interval [UI] 3·7 million to 4·8 million) deaths and 103·1 million (90·8 million 115·1 million)
British Columbia, Vancouver,
disability-adjusted life-years (DALYs) in 2015, representing 7·6% of total global deaths and 4·2% of global DALYs, BC, Canada (Prof M Brauer ScD);
59% of these in east and south Asia. Deaths attributable to ambient PM2·5 increased from 3·5 million (95% UI Health Canada, Ottawa, ON,
3·0 million to 4·0 million) in 1990 to 4·2 million (3·7 million to 4·8 million) in 2015. Exposure to ozone caused an Canada (R Burnett PhD,
H Shin PhD); St George’s,
additional 254 000 (95% UI 97 000–422 000) deaths and a loss of 4·1 million (1·6 million to 6·8 million) DALYs from
University of London,
chronic obstructive pulmonary disease in 2015. London, UK
(Prof H R Anderson MD);
Interpretation Ambient air pollution contributed substantially to the global burden of disease in 2015, which increased Institute for Health Metrics
and Evaluation, Seattle, WA,
over the past 25 years, due to population ageing, changes in non-communicable disease rates, and increasing air
USA (J Frostad MPH,
pollution in low-income and middle-income countries. Modest reductions in burden will occur in the most polluted K Estep MPA,
countries un­less PM2·5 values are decreased substantially, but there is potential for substantial health benefits from Prof L Dandona MD,
exposure reduction. G Freedman MPH,
Prof T Vos PhD,
Prof C J L Murray DPhil,
Funding Bill & Melinda Gates Foundation and Health Effects Institute. M H Forouzanfar PhD);
Sri Ramachandra University,
Copyright © The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Chennai, Tamil Nadu, India
(Prof K Balakrishnan PhD);
University of Utrecht, Utrecht,
Introduction Methods Netherlands
Exposure to ambient air pollution increases mortality Overview (Prof B Brunekreef PhD); Public
and morbidity and shortens life expectancy.1,2 The Global Attributing deaths and disability-adjusted life-years Health Foundation of India,
New Delhi, India
Burden of Diseases, Injuries, and Risk Factors Study (DALYs) to ambient air pollution requires spatially and (Prof L Dandona,
2015 (GBD 2015) estimated the burden of disease temporally resolved estimates of population-weighted R Dandona PhD); Auckland
attributable to 79 risk factors in 195 countries from exposure, specification of a theoretical minimum risk University of Technology,
1990 to 2015. GBD 2015 identified air pollution as a exposure level (TMREL), estimation of relative risks across Auckland, New Zealand
(Prof V Feigin PhD); United
leading cause of global disease burden, especially in low- the exposure distribution, and estimates of the deaths and States Environmental
income and middle-income countries.3 In view of the DALYs for diseases linked causally to air pollution. We Protection Agency,
important role of public policy in mitigating this risk and combined estimates of exposure and relative risk to Washington, DC, USA
(B Hubbell PhD); University of
the potential for substantial health benefits related to estimate the population-attributable fraction (PAF), the
Bath, Bath, UK (A Jobling PhD,
efforts to reduce emissions of climate-forcing agents, we proportion of deaths and DALYs attributable to exposure Prof G Shaddick PhD,
explored spatial and temporal trends in mortality and above the TMREL. The numbers of deaths and DALYs for M Thomas PhD); Fudan
burden of disease attributable to ambient air pollution specific diseases were multiplied by the PAF to estimate University, Yangpu Qu,
Shanghai, China
from 1990 to 2015 at global, regional, and country levels. the burden attributable to exposure. A more general

www.thelancet.com Vol 389 May 13, 2017 1907

 Articles
(Prof H Kan MD); University of
Queensland, St Lucia, QLD,
Australia (L Knibbs PhD); Emory
University, Atlanta, GA, USA
(Y Liu PhD); Dalhousie
University, Halifax, NS, Canada
(Prof R Martin PhD,
A van Donkelaar PhD);
Queensland University of
Technology, Brisbane, QLD,
Australia (L Morawska PhD);
Brigham Young University,
Provo, UT, USA
(Prof C A Pope III PhD);
International Agency for
Research on Cancer, Lyon,
France (K Straif PhD); and
European Commission,
Brussels, Belgium
(R van Dingenen PhD)
Correspondence to:
Dr Aaron J Cohen, Health Effects
Institute, Boston,
MA 02110-1817, USA
acohen@healtheffects.org
See Online for appendix
1908 www.thelancet.com Vol 389 May 13, 2017
Research in context
Evidence before this study
Literature reviews done by the US Environmental Protection
Agency, WHO, and others have shown that long-term exposure
to ambient air pollution increases mortality and morbidity from
cardiovascular and respiratory disease and lung cancer and
shortens life expectancy. Based on this evidence, the Global
Burden of Diseases, Injuries, and Risk Factors Study 2015
estimated the burden of disease attributable to 79 risk factors,
including ambient air pollution, in 195 countries and territories
from 1990 to 2015. The Global Burden of Diseases, Injuries,
and Risk Factors Study 2015 identified air pollution as a leading
cause of global disease burden, especially in low-income and
middle-income countries.
Added value of this study
In this Article, we show the crucial part played by broad
demographic and epidemiological trends. We estimated that
long-term exposure to ambient fine particle air pollution (PM2·5)
caused 4·2 million deaths and 103·1 million lost years of healthy
life in 2015, representing 7·6% of total global mortality, making it
the fifth-ranked global risk factor in 2015. Exposure to ozone was
responsible for an additional 254 000 deaths. Although global
rates of mortality due to PM2·5 exposure decreased from 1990 to
2015, the absolute numbers of attributable deaths and
disability-adjusted life-years increased because of rising levels of
pollution and increasing numbers of deaths from
non-communicable diseases in the largest low-income and
middle-income countries in east and south Asia, where
populations are growing and ageing. This research places the
burden of disease as a result of ambient air pollution within the
description of the methods used to estimate the PAF and
attributable burdens in GBD 2015 has been reported
previously;3 here, we present details specific to air pollution.
GBD 2015 extended the methods and datasets used in
GBD 2013 to estimate the burden of disease attributable to
ambient air pollution for 1990–2015.4,5 Estimates of
exposure to ambient air pollution were updated with
additional air pollution measurements and improved
estimation methods,3,6 relative risk estimates were updated
using more recent epidemiological studies and refined
statistical estimation techniques,3 and new methods were
developed and applied to identify key drivers of global
trends in the burden of disease attributable to ambient air
pollution. The GBD 2015 estimates update the entire time
series beginning in 1990, so the changes in methodology
affect the estimates for each year in the 1990–2015 interval
and supersede estimates previously released.
Estimation of exposure
Air pollution is a complex mixture of gases and particles
whose sources and composition vary spatially and
temporally. Population-weighted annual mean concen­
trations of particle mass with aerodynamic diameter less
context of other common potentially modifiable risk factors at a
national level, helping to prioritise air pollution from a population
health perspective. In the analysis of trends in the burden of
disease caused by ambient air pollution, we show where, and the
extent to which, ambient air pollution is changing as a contributor
to disease burden, and the extent to which the trends in the
attributable burden reflect progress, or absence of it, in reducing
exposure versus changes in demographic factors. In so doing, we
were able to elucidate the challenges that must be overcome to
reduce the public health effects of exposure to air pollution.
Implications of all the available evidence
Ambient air pollution was a leading risk factor for the global
burden of disease in 2015 and has remained stable, and its
contribution to the global burden of disease has remained
relatively stable, from 1990 to 2015. Trends in attributable
deaths reflect both demographic and epidemiological trends
and increasing levels of air pollution in low-income and
middle-income countries. Should these trends continue, they
will lead to increasing burdens if major reductions are not made
in pollution levels. Non-linear exposure–response functions
suggest modest reductions in burden in the most polluted
countries unless PM2·5 levels markedly decline. As a result, the
challenges for future reductions in the burden of disease
attributable to air pollution are substantial. International
experience has shown that exposure to ambient air pollution
and its associated burden of disease can be lowered for entire
populations via policy action at the national and subnational
levels via aggressive air quality management programmes,
focused on major sources of air pollution.
than 2·5 µm (PM2·5) and tropospheric ozone are the
two indicators used to quantify exposure to air pollution.
PM2·5 is the most consistent and robust predictor of
mortality in studies of long-term exposure.7,8 Ozone, a
gas produced via atmospheric reactions of precursor
emissions, is associated with respiratory disease
independent of PM2·5 exposure.9,10 We estimated exposure
to PM2·5 and ozone for the global population at spatial
scales relevant to human exposure (appendix pp 2–4).3,6
Global annual mean exposure to PM2·5 was estimated
in 5-year intervals from 1990 to 2015, at 0·1  × 
0·1°
(~11 km × 11 km at the equator) resolution using estimates
from satellites combined with a chemical transport
model, surface measure­ments, and geographical data.
We aggregated gridded exposure concentrations to
national-level population-weighted means using the
corresponding grid cell population value.11 National-level
population-weighted mean concentrations and the
95% uncertainty interval (95% UI) around this mean
were estimated by sampling 1000 draws of each grid cell
value and its uncertainty distribution.
As in previous assessments,4,5 we used a chemical
transport model to calculate a running 3-month mean
 Articles

(of daily 1 h maximum values) ozone concentration for each
grid cell over 1 year, from which we selected the maximum
of these values, consistent with epidemio­logical studies
that use a seasonal (summer) mean, while accounting for
global variation in the timing of the ozone (summer)
season. We estimated population-weighted mean ozone
concentrations and 95% UIs for each country as described
for PM2·5, assuming a normal distribution with a 95% UI
within 6% either side of the estimated mean concentration.
Theoretical minimum risk exposure level
TMREL was assigned a uniform distribution of
2·4–5·9 µg/m³ for PM2·5 and 33·3–41·9 parts per billion
for ozone, bounded by the minimum and fifth percentiles
of exposure distributions from outdoor air pollution
cohort studies (appendix pp 7, 11–14). The uniform
distribution represents the uncertainty regarding adverse
effects of low-level exposure.3,12,13
Risk estimation
We estimated the burden attributable to PM2·5 for
ischaemic heart disease (IHD), cerebrovascular disease
(ischaemic stroke and haemorrhagic stroke), lung cancer,
chronic obstructive pulmonary disease (COPD), and
lower respiratory infections (LRI), and the burden
attributable to ozone for COPD.3,14 Evidence linking these
diseases with exposure to ambient air pollution was
judged to be consistent with a causal relationship on
the basis of criteria specified for GBD risk factors.3
We developed integrated exposure–response functions
(IERs) for each cause of death to estimate the relative risk
of mortality over the entire global range of ambient
annual mean PM2·5 concentrations using risk estimates
from studies of ambient air pollution, household air
pollution, and second-hand smoke exposure and active
smoking (appendix pp 8–14).12,14 IERs assign con­
centrations of PM2·5 to each type of exposure on an
equivalent µg/m³ basis assuming that risk is determined
by the 24-h PM2·5 inhaled dose regardless of the exposure
source, consistent with previous findings.15,16 We updated
IERs from those used in GBD 2013 by adding additional
risk estimates for air pollution (appendix pp 11–14)
and active smoking.17 An alternative method to estimate
exposure to second-hand smoke was used that incor­
porated estimates of PM2·5 attributable to exposure per
cigarette, breathing rate, and number of cigarettes
smoked in the country where each study was done.
Further details are provided in the appendix (pp 8–14).

Ischaemic heart disease, age 25 years Ischaemic heart disease, age 50 years Ischaemic heart disease, age 80 years

3·0

2·5
Relative risk

2·0

1·5

1·0

Cerebrovascular disease, age 25 years Cerebrovascular disease, age 50 years Cerebrovascular disease, age 80 years

3·0
2·5
Relative risk

2·0

1·5

1·0

Lower respiratory infection, all ages Lung cancer, all ages Chronic obstructive pulmonary disease, all ages

3·0

2·5
Relative risk

2·0

1·5

1·0
0 25 50 75 100 125 0 25 50 75 100 125 0 25 50 75 100 125
PM2·5 (μg/m3) PM2·5 (μg/m3) PM2·5 (μg/m3)

Figure 1: Integrated exposure–response functions

Curves show the central estimate of the integrated exposure–response (solid lines) and their 95% uncertainty intervals (shaded areas). The relative risk equals 1 for
PM2·5 concentrations of 0–2·4 µg/m³ (ie, lower bound of the theoretical minimum risk exposure level uncertainty distribution). Additional details are provided in the
appendix (pp 7–15). PM2·5=particle mass with aerodynamic diameter less than 2·5 µm.

www.thelancet.com Vol 389 May 13, 2017 1909

 Articles
100
90
80
70
Population-weighted PM2·5 (ug/m3)
60
50
40
30
20
10
0
1990
Figure 2: Trends in population-weighted mean concentrations of particle mass with aerodynamic diameter less than 2·5 µm
Global data and data from the ten most populous countries are shown. Shaded areas are 95% uncertainty intervals. PM2·5=particle mass with aerodynamic diameter
less than 2·5 µm.
The IER has the mathematical form:
γ
–β(z – zcf ) +
IER(β,z)=1+ α× (1– e
where z is the level of PM2·5 and zcf is the TMREL, below
which no additional risk is assumed, with
(z – zcf )+=(z – zcf )
if z is greater than zcf and zero otherwise. Here, 1 + α is
the maximum risk, β is the ratio of the IER at low to high
concentrations, and γ is the power of PM2·5 concentration.
Epidemiological evidence suggests that the relative risks
for IHD and stroke decline with age.18 We modified the
particulate matter source-specific relative risk for both
IHD and stroke mortality as described by Burnett and
colleagues12 and applied this age modification to the relative
risks, fitting the IER model for each age group separately.
Observed relative risks were related to the IER within a
Bayesian framework using the STAN fitting algorithm,
as described in the appendix (pp 15–17). Given the true
values of the four parameters (α, β, γ, zcf), we assumed
that the logarithm of each study’s observed relative risk
was normally distributed, with mean defined by the IER
and variance given by the square of the observed SE of
the study-specific log-relative risk estimate plus an
additional variance term for each of the four sources on
PM2·5 exposure (outdoor air pollution, second-hand
smoke, household air pollution, and active smoking).
Details regarding model fitting and code are provided in
the appendix (pp 15–17).
1910 www.thelancet.com Vol 389 May 13, 2017
Bangladesh
India
Pakistan
China
Global
Nigeria
Russia
Indonesia
Japan
Brazil
USA
1995 2000 2005 2010 2015
Year
We calculated 1000 predicted values of the IER for each
PM2·5 concentration based on the posterior distributions
) of (α, β, γ) and the prespecified uniform distribution of
TMREL to characterise uncertainty in the estimates of
the IER. The mean of the 1000 IER predictions at each
concentration was used as the central estimate, with
uncertainty defined by 95% UIs.
We estimated the relative risk of COPD mortality from
ozone exposure using a linear exposure–response function
for respiratory mortality from Jerrett and colleagues.10
Additional details are provided in the appendix (p 14).
Estimation of PAF and burden
We calculated DALYs and deaths attributable to ambient
air pollution by applying the year-specific, location-
specific, age-specific, and sex-specific PAF to the
numbers of DALYs and deaths, as described in detail
elsewhere.19,20 PAF estimation methods are summarised
in the appendix (pp 20–21).
Role of the funding source
The funders of the study had no role in study design,
data collection, data analysis, data interpretation, or
writing of the report. All authors had full access to all the
data in the study and AJC and MB had final responsibility
for the decision to submit for publication.
Results
Figure 1 shows IERs for the five causes of death. The
functions are all non-linear, with a greater change in
relative risk for lower concentrations compared with
 Articles

higher values. We fit age-specific functions for IHD

A
and cerebrovascular disease, and estimated decreasing 1990 rank 2015 rank
relative risks as age increased from 25 years to 80 years.
1 High systolic blood pressure 1 High systolic blood pressure
Global population-weighted PM2·5 increased by 11·2%
2 Smoking 2 Smoking
from 1990 (39·7 µg/m³) to 2015 (44·2 µg/m³), increasing
3 Childhood undernutrition 3 High fasting plasma glucose
most rapidly from 2010 to 2015 (figure 2). Among the
4 Ambient particulate matter pollution 4 High total cholesterol
world’s ten most populous countries, exposures since 2010
5 Household air pollution from solid fuels 5 Ambient particulate matter pollution
increased in Bangladesh and India and were stable but
6 High total cholesterol 6 Diet high in sodium
remained high in Pakistan and China. Exposures decreased
7 High fasting plasma glucose 7 High body-mass index
substantially in Nigeria and were low and slightly decreased
8 Diet high in sodium 8 Diet low in whole grains
in the USA, Brazil, and Russia. Population-weighted
concentrations were low and stable in Japan and Indonesia. 9 High body-mass index 9 Diet low in fruits

The highest estimated population-weighted mean con­ 10 Unsafe water source 10 Household air pollution from solid fuels

centrations in 2015 were for Qatar (107·3 µg/m³), 11 Diet low in whole grains 11 Impaired kidney function

Saudi Arabia (106·2 µg/m³), and Egypt (104·7 µg/m³), 12 Diet low in fruits 12 Alcohol use
followed by Bangladesh (89·4 µg/m³), Mauritania 13 Alcohol use 13 Diet low in nuts and seeds
(85·1 µg/m³), Libya (79·2 µg/m³), Nepal (75·0 µg/m³), 14 Unsafe sanitation 14 Diet low in vegetables
and India (74·3 µg/m³). The population-weighted mean 15 No handwashing with soap 15 Low physical activity
PM2·5 in China was 58·4 µg/m³, with provincial 16 Diet low in vegetables 16 Diet low in seafood omega3 fatty acids
population-weighted means ranging from 19·1 µg/m³ to 17 Impaired kidney function 17 Unsafe sex
79·3 µg/m³. The lowest estimated population-weighted 18 Diet low in nuts and seeds 18 Childhood undernutrition
means were in several Pacific island nations and 19 Suboptimal breastfeeding 19 Unsafe water source
territories (Kiribati, American Samoa, Samoa, Tonga, 20 Low physical activity 20 No handwashing with soap
Solomon Islands, Fiji, and Guam), Brunei, Sweden, 21 Diet low in seafood omega3 fatty acids 21 Second-hand smoke
Greenland, New Zealand, Australia, Finland, Iceland, 22 Second-hand smoke 22 Unsafe sanitation
Liberia, and Canada (all ≤8·0 µg/m³). 23 Unsafe sex 23 Diet high in processed meat
Population-weighted ozone levels increased by 7·2% 24 Diet high in processed meat 24 Suboptimal breastfeeding
globally from 1990 (56·8 parts per billion [ppb]) to 2015
(60·9 ppb). Within the world’s ten most populous B
countries, increases of 14–25% were noted in China, 1 Childhood undernutrition 1 High systolic blood pressure
India, Pakistan, Bangladesh, and Brazil, with smaller 2 Unsafe water source 2 Smoking
increases in Japan and negligible changes in Russia and 3 High systolic blood pressure 3 High fasting plasma glucose
Nigeria (data not shown). Decreases in population- 4 Household air pollution from solid fuels 4 High body-mass index
weighted concentrations were noted in the USA (5·2%; 5 Smoking 5 Childhood undernutrition
from 70·2 ppb to 66·5 ppb) and Indonesia (12·9%; from 6 Ambient particulate matter pollution 6 Ambient particulate matter pollution
50·2 ppb to 43·7 ppb). 7 Unsafe sanitation 7 High total cholesterol
Long-term exposure to PM2·5 contributed to 4·2 million 8 Suboptimal breastfeeding 8 Household air pollution from solid fuels
(95% UI 3·7 million to 4·8 million) deaths and to a loss 9 No handwashing with soap 9 Alcohol use
of 103·1 million (90·8 million to 115·1 million) DALYs in
10 High fasting plasma glucose 10 Diet high in sodium
2015, representing 7·6% of total global deaths and 4·2%
11 Alcohol use 11 Diet low in whole grains
of global DALYs, which is an increase from 1990. In 2015,
12 High total cholesterol 12 Unsafe sex
ambient PM2·5 was the fifth-ranked risk factor for global
13 High body-mass index 13 Diet low in fruits
deaths and sixth-ranked risk factor for DALYs among the
14 Diet high in sodium 14 Unsafe water source
risk factors included in GBD 2015 (figure 3). DALYs
15 Diet low in whole grains 15 Impaired kidney function
attributable to long-term exposure to PM2·5 consisted of
16 Diet low in fruits 16 Iron deficiency
99·2 million (95% UI 87·7 million to 111·0 million) years
17 Iron deficiency 17 Diet low in nuts and seeds
of life lost and 3·9 million (2·6 million to 5·2 million)
years lived with disability in 2015. 18 Second-hand smoke 18 No handwashing with soap

Household air pollution from solid fuel use was 19 Vitamin A deficiency 19 Unsafe sanitation

responsible for 2·8 million (95% UI 2·2 million to 20 Unsafe sex 20 Diet low in vegetables

3·6 million) deaths and 85·6 million (66·7 million to 21 Impaired kidney function 21 Low physical activity
22 Diet low in vegetables 22 Suboptimal breastfeeding
23 Diet low in nuts and seeds 23 Second-hand smoke
Figure 3: Leading level 3 Global Burden of Diseases global risk factors for 24 Low physical activity 24 Vitamin A deficiency
deaths (A) and disability-adjusted life-years (B), 1990 and 2015
Behavioural risks
Risks are connected by lines between years; solid lines show risks that have
Environmental or occupational risks
stayed the same or moved higher in the ranking and dashed lines show risks that Metabolic risks
have moved lower.

www.thelancet.com Vol 389 May 13, 2017 1911

 Articles

106·1 million) DALYs in 2015. Together, ambient and
household air pollution were estimated to have caused
6·4 million (5·7 million to 7·3 million) deaths in 2015.
Mortality from cardiovascular disease (IHD and
cerebrovascular disease) accounted for most deaths and
DALYs attributable to ambient PM2·5 air pollution
(figure 4; table 1). Ambient PM2·5 air pollution contributed
to 17·1% of IHD, 14·2% of cerebrovascular disease,
16·5% of lung cancer, 24·7% of LRI, and 27·1% of COPD
mortality in 2015 according to GBD compare.21
Age-standardised death and DALY rates due to
exposure to PM2·5 were higher in males than females
(table 1), as a result of higher all-cause mortality rates in
males (1018·6 per 100  000 males vs 703·4 per
100 000 females21). They were also higher in elderly
4·5
4·0
Deaths attributable to PM2·5 (millions)
people (age >70 years) than in children (age <5 years;
table 1), mainly because of age-related differences in
mortality from non-communicable diseases (41·4 per
100 000 children aged 1–5 years vs 2914·4 per
100 000 adults aged 70–74 years21). Ambient PM2·5
contributed to 202 000 (95% UI 152 700–254 600) deaths
and 17·4 million (13·1 million to 21·9 million) DALYs
from LRI in children younger than 5 years.
Deaths attributable to long-term exposure to PM2·5 in
2015 varied substantially among countries (figure 5).
South and east Asia contributed 59% of the 4·2 million
global deaths attributable to ambient PM2·5 in 2015
(1·36 million deaths [95% UI 1·19 million to 1·56 million]
in south Asia and 1·14 million deaths [0·97 million to
1·31 million] in east Asia). In World Bank high-income
countries, exposure to ambient PM2·5 contributed to 4·3%
of total deaths in 2015 versus 9·0% in upper-middle-
income, 8·7% in lower-middle-income, and 4·9% in low-
income countries. These differences in attributable
mortality mostly reflect the fraction of total deaths from

3·5
cardiovascular disease among countries.3 The highest
3·0
age-standardised rates of death due to PM2·5 exposure
2·5 were in southern Asia (133·4 per 100 000 population,
95% UI 114·2–152·6), central Asia (85·2 per
2·0
100 000 population, 72·0–98·9), and eastern Asia
1·5 (83·2 per 100 000 population, 70·4–95·6). Rates in high-
income North American (USA, Canada, and Greenland;
1·0
17·8 per 100 000 people [95% UI 13·6–22·9]), Asian
0·5 (18·7 per 100  000 people [14·6–23·7]), and western
European countries (19·9 per 100 000 [15·9–24·8]) were
0
1990 1995 2000 2005 2010 2015 four to eight times lower (appendix pp 26–1078).
Year Table 2 provides 2015 mortality and DALY estimates for
Tracheal, bronchial, and lung cancer Cerebrovascular disease Lower respiratory
the world’s ten most populous countries in 2015.
Ischaemic heart disease Chronic obstructive pulmonary disease infections
Ambient PM2·5 ranked among the top ten risk factors for
Figure 4: Deaths attributable to ambient particulate matter pollution by year and cause mortality in each of the world’s most populous countries.
PM2·5=particle mass with aerodynamic diameter less than 2·5 µm. China and India combined had the largest numbers of

Deaths, in thousands (95% UI) Age-standardised deaths DALYs, in thousands (95% UI) Age-standardised DALYs
per 100 000 people per 100 000 people
(95% UI) (95% UI)
All causes 4241·1 (3698·0–4776·7) 66·0 (57·2–74·8) 103 066·2 (90 829·6–115 072·6) 1490·9 (1312·4–1665·6)
Disease
Lower respiratory infection 675·0 (491·9–889·0) 10·1 (7·4–13·4) 28 359·9 (21 141·8–35 796·9) 390·9 (290·9–494·3)
Lung cancer 283·3 (178·4–398·7) 4·4 (2·7–6·1) 6209·1 (3934·9–8689·3) 90·9 (57·5–127·3)
Ischaemic heart disease 1521·1 (1231·7–1821·2) 23·6 (18·9–28·5) 32 406·0 (27 078·2–37 427·4) 470·7 (394·6–543·0)
Cerebrovascular disease 898·1 (717·6–1083·6) 14·0 (11·0–17·1) 19 242·8 (16 095·9–22 679·7) 281·2 (234·4–331·4)
Chronic obstructive 863·6 (538·5–1212·8) 14·0 (8·7–19·6) 16 848·2 (10 517·4–23 590·0) 257·2 (160·3–360·6)
pulmonary disease
Sex
Male 2455·4 (2140·2–2752·9) 83·9 (72·5–94·7) 62 894·7 (55 545·7–70 098·2) 1888·8 (1659·4–2113·6)
Female 1785·7 (1546·2–2049·2) 50·8 (44·0–58·4) 40 171·5 (35 205·5–45 382·8) 1127·4 (986·6–1275·4)
Age
Children <5 years 202·6 (152·7–254·6) 30·1 (22·7–37·8) 17 431·1 (13 139·7–21 906·3) 2585·9 (1949·1–3249·5)
Elderly >70 years 2228·3 (1842·0–2653·9) 562·7 (465·1–670·8) 25 073·0 (20 775·2–29 511·1) 6302·2 (5226·3–7419·8)

DALY=disability-adjusted life-year. UI=uncertainty interval.

Table 1: Global deaths, disability-adjusted life-years, and age-standardised rates attributable to ambient particulate matter pollution in 2015

1912 www.thelancet.com Vol 389 May 13, 2017


Deaths (%)
No data
<3·0
3·0–3·5
3·6–4·0
4·1–4·4
4·5–4·8
4·9–5·5
5·6–6·7
6·8–7·3
7·4–8·5
≥8·6
ATG
Dominica
Caribbean LCA
Figure 5: Deaths attributable to ambient particulate matter pollution in 2015
ATG=Antigua and Barbuda. FSM=Federated States of Micronesia. Isl=Island. LCA=Saint Lucia. TLS=Timor-Leste. TTO=Trinidad and Tobago. VCT=Saint Vincent and the Grenadines.
attributable deaths and DALYs: 52% and 50% of the
respective global totals. Pakistan, India, and Bangladesh
had the highest age-adjusted mortality rates, more than
seven times higher than those of Japan and the USA
(table 2; appendix pp 26–1078).
Global mortality due to ambient PM2·5 increased from
1990 to 2015. Attributable deaths rose from 3·5 million
(95% UI 3·0 million to 4·0 million) in 1990 to 3·8 million
(3·3 million to 4·3 million) in 2000, and 4·2 million
(3·7 million to 4·8 million) in 2015 (figure 4). However,
age-standardised PM2·5 mortality rates decreased from
65·6 per 100 000 people (95% UI 56·9–74·9) in 1990 to
57·5 per 100 000 people (50·2–64·8) in 2015.
Trends in PM2·5-attributable mortality among countries
largely reflect changes in PM2·5-attributable mortality from
cardiovascular disease (appendix pp 24, 26–559). In World
Bank high-income countries, the all-age proportion of
PM2·5-attributable cardiovascular disease deaths decreased
from 10·0% to 8·1% as a result of reductions in
cardiovascular mortality and decreasing levels of PM2·5. By
contrast, in World Bank low-income countries, it increased
from 13·1% to 13·2%, and in lower-middle-income
countries from 15·9% to 16·5%, between 1990 and 2015.
Trends in PM2·5-attributable mortality at the global and
national levels reflect the influence not only of changing
air quality, but also of demography and underlying
mortality rates. We calculated the contribution of changes
www.thelancet.com Vol 389 May 13, 2017 1913
Articles
VCT Barbados Comoros Marshall Isl Kiribati
West Africa Eastern
Mediterranean
Solomon Isl FSM
Grenada Maldives Mauritius Malta
Vanuatu Samoa
TTO TLS Seychelles Persian Gulf Singapore Balkan Peninsula Fiji Tonga
in each of four factors—population growth, population
ageing, age-standardised rates of mortality (IHD,
cerebrovascular disease, COPD, lung cancer, and LRI),
and exposure to ambient PM2·5—to the net change in
mortality attributable to ambient PM2·5 between 1990 and
2015 globally and for the ten most populous countries
(appendix pp 22–23). Figure 6 shows the changes in
mortality attributable to ambient PM2·5 from 1990 to 2015
according to the contributions of these four factors. Age-
standardised mortality decreased in all ten countries, with
Nigeria, Russia, Brazil, Indonesia, Pakistan, and the USA
also experiencing decreases in exposure. These decreases
were offset by increases in population growth and
population ageing in most countries. Consequently, net
increases in attributable mortality were noted in all
countries except Nigeria and the USA. In China, India,
Bangladesh, and Japan, increases in exposure combined
with increases in population growth and ageing resulted
in net increases in attributable mortality. In Brazil,
Russia, Indonesia, and Pakistan, despite decreasing
exposure, population growth (except in the case of Russia)
and the ageing of the population led to a net increase in
attributable mortality. In the USA, reductions in exposure
offset increases in population and ageing, leading to a net
decrease in attributable burden.
GBD 201519 estimated disease burden and mortality
rates at the subnational level for China, the USA, and
 Articles

Deaths, in thousands Risk factor rank Deaths per 100 000 people DALYs, in thousands (95% UI) DALYs per 100 000 people Population-weighted
(95% UI) for deaths (95% UI) (95% UI) mean PM2·5 (µg/m³; 95% UI)
China 1108·1 (948·7–1272·8) 1 84·3 (71·5–96·7) 21 778·7 (18 903·5–24 584·2) 1478·6 (1275·9–1675·6) 58·4 (58·1–58·7)
India 1090·4 (936·6–1254·8) 2 133·5 (112·8–154·9) 29 609·6 (25 923·3–33 562·7) 2922·1 (2527·3–3327·5) 74·3 (73·9–74·8)
USA 88·4 (66·8–115·0) 6 18·5 (14·2–23·7) 1485·9 (1166·3–1841·7) 337·1 (265·0–416·8) 8·4 (8·4–8·5)
Indonesia 78·6 (62·0–96·7) 7 49·9 (38·5–61·6) 2185·0 (1730·4–2716·2) 1081·1 (860·4–1324·2) 15·4 (15·1–15·7)
Brazil 52·3 (41·9–65·1) 9 30·9 (24·2–39·0) 1083·9 (884·0–1322·7) 573·7 (467·3–702·3) 11·4 (11·2–11·5)
Pakistan 135·1 (114·3–159·2) 4 136·3 (113·7–163·5) 4217·3 (3545·1–4916·3) 3114·2 (2651·3–3657·7) 65·0 (63·8–66·2)
Nigeria 50·9 (35·7–73·2) 10 68·9 (48·5–101·7) 2410·0 (1640·4–3387·0) 1581·0 (1107·6–2237·2) 38·0 (37·5–38·5)
Bangladesh 122·4 (103·2–144·4) 5 133·2 (111·8–158·4) 3408·0 (2920·3–3945·8) 2972·0 (2533·4–3469·1) 89·4 (87·3–91·7)
Russia 136·9 (111·3–161·1) 3 62·6 (51·8–73·2) 2601·6 (2194·8–3007·2) 1255·0 (1077·8–1431·1) 16·6 (16·2–17·0)
Japan 60·6 (44·5–81·4) 8 16·8 (12·8–21·9) 705·8 (561·2–891·0) 261·7 (212·8–319·2) 13·3 (13·1–13·6)

Countries are shown in order of population size in 2015. DALY=disability-adjusted life-year. PM2·5=particle mass with aerodynamic diameter less than 2·5 µm. UI=uncertainty interval.

Table 2: 2015 estimates of mortality and disability-adjusted life-years attributable to ambient particulate matter pollution and population-weighted mean particulate matter pollution
in the world’s ten most populous countries

150 Population growth Exposure to ozone contributed to 254 000 (95% UI

Population ageing 97 000–422 000) deaths globally and a loss of 4·1 million
Age-standardised
(1·6 million to 6·8 million) DALYs from COPD in 2015.
mortality
Change in mortality attributable to ambient PM2·5 (%)

Exposure In 2015, ambient ozone was the 34th-ranked risk factor

100
Net change for global deaths and 42nd-ranked risk factor for DALYs
among the 79 risk factors assessed in GBD 2015.
Exposure to ozone contributed to an estimated 8·0%
50
(95% UI 3·0–13.3) of global COPD mortality in 2015,
with China, India, and the USA experiencing some of the
highest mortality rates (figure 7A; appendix pp 1079–1362).
0 The ozone-attributable COPD mortality rate increased in
many countries from 1990 to 2015. Global deaths and
DALYs attributable to ozone exposure increased from
–50 1990 to 2015, as a result of increases in both levels of
ozone and COPD mortality (figure 7B; appendix
pp 1079–1362).
–100
Discussion
n
al

sia

an

il

sia

ria
di

in

az
ta

es

US
ob

Jap

ge
ne

s
In

Ch

Br
kis

lad

Ru
Gl

Ni
do

In this Article, we present, to our knowledge, the most

Pa

ng

In
Ba

comprehensive assessment so far of the global status

Figure 6: Changes in mortality attributable to ambient particulate matter pollution according to
population-level determinants by country from 1990 to 2015
PM2·5=particle mass with aerodynamic diameter less than 2·5 µm.
several other countries. When these data were combined
with PM2·5 exposures estimated globally at fine (0·1 × 0·1°)
resolution, we were able to estimate subnational burden
attributable to PM2·5 exposure. In China, where ambient
PM2·5 contributed to 1·1 million (95% UI 1·0 million
to 1·8 million) deaths in 2015, the provincial-level
PM2·5-attributable age-standardised rates varied by more
than three times, from 132·1 deaths per 100 000 people
(95% UI 97·6–172·0) in Qinghai to 40·6 deaths per
100 000 people (30·2–50·4) in Hong Kong. In the USA,
where ambient PM2·5 contributed to 88 400 (95% UI
66 800–115 000) deaths in 2015, state-level PM2·5-attributable
age-standardised death rates also varied by about three
times, from 27·1 deaths per 100 000 (95% UI 21·2–34·1)
in Mississippi to 8·1 deaths per 100  000 (5·1–11·7)
in Hawaii.
and trends in the burden of disease attributable to
ambient air pollution, highlighting the crucial part
played by broad demographic and epidemiological
trends. Estimation of spatially resolved trends over a
25-year period and assessment of the contributions of
exposure and epidemiological and demographic factors
to the attributable disease burden provides important
insights for the development of policies to reduce the
health effects of air pollution. These estimates are made
in the context of assessment of burden attributable to
other risk factors, allowing for direct comparisons and
priority setting.
We estimated that long-term exposure to ambient
PM2·5 caused 4·2 million deaths and 103·1 million lost
years of healthy life in 2015, and exposure to ozone
caused an additional 254 000 deaths. PM2·5 caused an
estimated 7·6% of total global mortality in 2015 and was
the fifth-ranking global mortality risk factor. Although
global rates of mortality due to PM2·5 exposure decreased

1914 www.thelancet.com Vol 389 May 13, 2017

 Articles

A Deaths per 100000 people in 2015

Deat hs per 100000 people (%)

Nodata
<0·10
0·10-0·19
0·20-0·29
0'30-0049
0·50-0·79
0·80-0·99
1·00-1·39
1040-1·89
1·90-2·59
,2·60

Barbados [] ~ rn
•
WestAfrica Eastern

Caribbean

B Percentage change in deaths from 1990to 2015

Change in deaths from 1990 to 2015 (%)

Nodata
<-37·7
-37·7 to -26,3
-2604to -1404
-14·5 to 004.
0·5 to 13·6
13·7 to 3004
30·5 to 58·0
58·1 to 92·3
9204 to 207·3
,20704

Barbados

IGr~n:da I
EJ El
Figure 7: Proportion of deaths attributable to ozone (A) in 2015 and percentage change from 1990 (B)
ATG=Antigua and Barbuda. FSM=Federated States of Micronesia. Isl=Island. LCA=Saint Lucia. TLS=Timor-Leste. TTO=Trinidad and Tobago. VCT=Saint Vincent and the Grenadines.

www.thelancet.com Vol 389 May 13, 2017 1915

 Articles
from 1990 to 2015 as a result of improved air quality in
high-income countries and declining mortality rates for
cardiovascular diseases, the absolute numbers of
attributable deaths and DALYs increased as a result of
increases in pollution and the absolute numbers
of deaths from non-communicable diseases, especially
in China and India, where populations are both growing
and ageing. Household air pollution from the burning
of solid fuels is also a major cause of mortality in low-
income and middle-income countries, and together
with ambient air pollution poses a substantial public
health challenge.
Absolute numbers of attributable deaths and DALYs
were higher in GBD 20153 than estimated in GBD 2013.4
These differences are mainly a result of changes in the
underlying disease burden estimates3 and to updates to
the IER (appendix pp 8–14), which estimated higher
relative risks in 2015 than in 2013 for IHD,
cerebrovascular disease, LRI, COPD, and lung cancer.
National-level population-weighted exposure estimates
also increased between GBD 2013 and GBD 2015
(appendix pp 2–6). Because of the updated data and
methods described earlier, we consider the current
estimates to be more accurate.
Our results assume that the toxicity of ambient PM2·5
depends only on the magnitude of concentration, but
not on the source, such as coal burning or vehicular
emissions, or chemical composition, which vary among
and within countries.22,23 However, despite substantial
effort, neither epidemiological nor toxicological research
has identified particular sources or components that
uniquely determine the toxicity of the PM2·5 mixture,
and therefore the evidence does not support the
development and application of source-specific relative
risk functions for burden estimation.15,16 This issue
remains an active area of research and is a source of
uncertainty in our estimates.
In the past few years, other researchers have estimated
the burden of disease due to air pollution using different
data and methods. Recent estimates from WHO24 of
3·0 million deaths in 2012 used the same exposure
estimates as presented here, but an earlier (GBD 2013)
version of the IER and somewhat different baseline
disease burden estimates. Lelieveld and colleagues25
analysed source sector contributions to air pollution and
the resulting disease burden in 2010 and estimated the
burden in 2050. These estimates used an older (GBD 2010)
IER. Furthermore, the coarse spatial resolution
(~100 × 100 km) of the exposure estimates introduced
errors via spatial misalignment between exposure and
population density compared with our estimates.
As in any assessment of this scope, this study has
limitations. Since the GBD will be regularly updated, we
anticipate enhancements to the methodology in the
future to address them. First, we have probably
underestimated the complete burden of disease
attributable to air pollution. Although the causes of
1916 www.thelancet.com Vol 389 May 13, 2017
mortality we included make up four of the five leading
global causes of death in 2015,3 findings from systematic
reviews in the past 10 years have shown that PM2·5
exposure is also associated with low birthweight and
preterm birth,26 asthma,9 and type 2 diabetes.27 Future
updates of GBD estimates will consider these other
causes of mortality and morbidity should they meet GBD
inclusion criteria.
Second, our estimate of the importance of ambient
PM2·5 assumes that exposure does not affect the
prevalence of other mortality risk factors. However, if
long-term exposure to PM2·5 causes high blood pressure,
then some amount of the PM2·5 burden would be
mediated by its effect on high blood pressure. Mediation
analysis was used in GBD 2015 to more accurately
apportion the burden attributable to other risk factors
such as diet and high blood pressure, but an absence of
longitudinal studies precludes such analyses for
ambient PM2·5.3
Third, because large-scale cohort studies of PM2·5 and
mortality are absent in the most polluted countries, the
IERs were developed to estimate the effects of exposure
at levels above those observed in air pollution cohort
studies done in the USA, Canada, and western Europe,
but the magnitude of the excess relative risk from PM2·5
exposure at high levels of PM2·5 remains uncertain. In
Chinese cohort studies from the past 5 years, other
metrics were used, such as total suspended particles
and PM10,28,29 and findings from a few analyses that
converted these metrics to PM2·5 suggest that the IERs
provide reasonable estimates of effects at high levels of
ambient pollution.12,30
Fourth, although we included estimates of the effect of
seasonal ozone exposure on COPD mortality, less
evidence is available for this relationship than that
linking PM2·5 with COPD or the other causes of mortality.
However, a causal link between increased COPD
mortality and long-term exposure to ozone is, in our
view, supported by a large body of evidence linking ozone
exposure mortality to adverse effects on the respiratory
system, including chronic changes in lung structure and
function in human beings and non-human primates,
and increased morbidity and mortality from COPD due
to short-term and long-term exposure, especially in the
warmer seasons.10,31,32
In conclusion, ambient air pollution contributes
substantially to the global burden of disease, which
has increased over the past 25 years, as a result of
both demographic and epidemiological trends and
increasing levels of air pollution in low-income and
middle-income countries. Should these trends continue,
major reductions in pollution levels will be needed to
avoid increases in disease burden. Moreover, the non-
linear IERs imply modest reductions in burden in the
most polluted countries unless PM2·5 concentrations
decline markedly.3,33,34 As a result, the challenges for
future reductions in the burden of disease attributable to

air pollution are substantial. For example, using earlier
attributable burden estimates and future mortality
predictions, Apte and colleagues33 estimated that air
pollution levels in 2030 in China would need to decline
by 29%, and those in India by 20%, to maintain
per-person mortality at 2010 levels, although the
economic35 and public health benefits of even
incremental reductions would probably be substantial in
view of the large populations affected.34 Exposure to
ambient air pollution and its associated burden of
disease can potentially be lowered for entire populations
via policy action at the national and subnational levels.
As the experience in the USA suggests,36 changes in
ambient PM2·5 associated with aggressive air quality
management programmes, focused on major sources23
of air pollution including coal combustion, household
burning of solid fuels, and road transport, can lead to
increased life expectancy over short timeframes.
Contributors
AJC, MB, RB, HRA, and MHF prepared the first draft. AJC, MB, and RB
finalised the draft based on comments from other authors and reviewer
feedback. All authors provided data, developed models, analysed data,
reviewed results, provided guidance on methodology, or reviewed the
manuscript, or did a combination thereof.
Declaration of interests
We declare no competing interests.
References
1 United States Environmental Protection Agency. 2009 final report:
integrated science assessment for particulate matter. 2009.
http://cfpub.epa.gov/ncea/risk/recordisplay.cfm?deid=216546
(accessed Dec 22, 2015).
2 WHO. Air quality guidelines—global update 2005. http://www.who.
int/phe/health_topics/outdoorair/outdoorair_aqg/en/ (accessed
Dec 22, 2015).
3 GBD 2015 Risk Factors Collaborators. Global, regional, and national
comparative risk assessment of 79 behavioural, environmental and
occupational, and metabolic risks or clusters of risks, 1990–2015:
a systematic analysis for the Global Burden of Disease Study 2015.
Lancet 2016; 388: 1659–724.
4 GBD 2013 Risk Factors Collaborators, Forouzanfar MH,
Alexander L, et al. Global, regional, and national comparative risk
assessment of 79 behavioural, environmental and occupational,
and metabolic risks or clusters of risks in 188 countries, 1990–2013:
a systematic analysis for the Global Burden of Disease Study 2013.
Lancet 2015; 386: 2287–323.
5 Brauer M, Freedman G, Frostad J, et al. Ambient air pollution
exposure estimation for the Global Burden of Disease 2013. Environ
Sci Technol 2016; 50: 79–88.
6 Shaddick G, Thomas ML, Jobling A, et al. Data integration model
for air quality: a hierarchical approach to the global estimation of
exposures to ambient air pollution. ArXiv1609.00141 Stat 2016;
published online Sept 1. http://arxiv.org/abs/1609.00141 (accessed
Nov 4, 2016).
7 Chen H, Goldberg MS, Villeneuve PJ. A systematic review of the
relation between long-term exposure to ambient air pollution and
chronic diseases. Rev Environ Health 2008; 23: 243–97.
8 Hoek G, Krishnan RM, Beelen R, et al. Long-term air pollution
exposure and cardio-respiratory mortality: a review.
Environ Health Glob Access Sci Source 2013; 12: 43.
9 United States Environmental Protection Agency. Integrated science
assessment of ozone and related photochemical oxidants (second
external review draft). Washington, DC: US Environmental
Protection Agency, 2011. http://cfpub.epa.gov/ncea/isa/recordisplay.
cfm?deid=226363 (accessed March 3, 2017).
10 Jerrett M, Burnett RT, Pope CA 3rd, et al. Long-term ozone
exposure and mortality. N Engl J Med 2009; 360: 1085–95.
www.thelancet.com Vol 389 May 13, 2017 1917
Articles
11 Center for International Earth Science Information Network—
CIESIN—Columbia University. 2016. Gridded population of the
world, version 4 (GPWv4): population count. Palisades, NY: NASA
Socioeconomic Data and Applications Center (SEDAC). http://dx.doi.
org/10.7927/H4X63JVC (accessed March 16, 2017).
12 Burnett RT, Pope CA III, Ezzati M, et al. An integrated risk function
for estimating the global burden of disease attributable to ambient
fine particulate matter exposure. Environ Health Perspect 2014;
122: 397–403.
13 Lim SS, Vos T, Flaxman AD, et al. A comparative risk assessment of
burden of disease and injury attributable to 67 risk factors and risk
factor clusters in 21 regions, 1990–2010: a systematic analysis for the
Global Burden of Disease Study 2010. Lancet 2012; 380: 2224–60.
14 Shin HH, Cohen AJ, Pope CA, et al. Meta-analysis methods to
estimate the shape and uncertainty in the association between
long-term exposure to ambient fine particulate matter and
cause-specific mortality over the global concentration range.
Risk Anal Off Publ Soc Risk Anal 2016; 36: 1813–25.
15 Lippmann M, Chen L-C, Gordon T, Ito K, Thurston GD.
National Particle Component Toxicity (NPACT) Initiative: integrated
epidemiologic and toxicologic studies of the health effects of
particulate matter components. Res Rep Health Eff Inst 2013;
177: 5–13.
16 Stanek LW, Sacks JD, Dutton SJ, Dubois J-JB. Attributing health
effects to apportioned components and sources of particulate matter:
an evaluation of collective results. Atmos Environ 2011; 45: 5655–63.
17 Doll R, Peto R, Wheatley K, Gray R, Sutherland I. Mortality in
relation to smoking: 40 years’ observations on male British doctors.
BMJ 1994; 309: 901–11.
18 Singh GM, Danaei G, Farzadfar F, et al. The age-specific
quantitative effects of metabolic risk factors on cardiovascular
diseases and diabetes: a pooled analysis. PLoS One 2013;
8: e65174.
19 GBD 2015 Mortality and Causes of Death Collaborators.
Global, regional, and national life expectancy, all-cause mortality,
and cause-specific mortality for 249 causes of death, 1980–2015:
a systematic analysis for the Global Burden of Disease Study 2015.
Lancet 2016; 388: 1459–544.
20 GBD 2015 DALYs and HALE Collaborators. Global, regional, and
national disability-adjusted life-years (DALYs) for 315 diseases and
injuries and healthy life expectancy (HALE), 1990–2015: a systematic
analysis for the Global Burden of Disease Study 2015. Lancet 2016;
388: 1603–58.
21 Institute for Health Metrics and Evaluation. GBD compare.
https://vizhub.healthdata.org/gbd-compare (accessed March 1, 2017).
22 Thurston GD, Burnett RT, Turner MC, et al. Ischemic heart disease
mortality and long-term exposure to source-related components of
U.S. fine particle air pollution. Environ Health Perspect 2016;
124: 785–94.
23 GBD MAPS Working Group. Burden of disease attributable to
coal-burning and other air pollution sources in China. Aug 20, 2016.
https://www.healtheffects.org/publication/burden-disease-
attributable-coal-burning-and-other-air-pollution-sources-china
(accessed Nov 2, 2016).
24 WHO. WHO global urban ambient air pollution database
(update 2016). http://www.who.int/phe/health_topics/outdoorair/
databases/cities/en/ (accessed March 3, 2017).
25 Lelieveld J, Evans JS, Fnais M, Giannadaki D, Pozzer A.
The contribution of outdoor air pollution sources to premature
mortality on a global scale. Nature 2015; 525: 367–71.
26 Stieb DM, Chen L, Eshoul M, Judek S. Ambient air pollution, birth
weight and preterm birth: a systematic review and meta-analysis.
Environ Res 2012; 117: 100–11.
27 Wang B, Xu D, Jing Z, Liu D, Yan S, Wang Y. Effect of long-term
exposure to air pollution on type 2 diabetes mellitus risk: a systemic
review and meta-analysis of cohort studies. Eur J Endocrinol 2014;
171: R173–82.
28 Cao J, Yang C, Li J, et al. Association between long-term exposure to
outdoor air pollution and mortality in China: a cohort study.
J Hazard Mater 2011; 186: 1594–600.
29 Zhou M, Liu Y, Wang L, Kuang X, Xu X, Kan H. Particulate air
pollution and mortality in a cohort of Chinese men.
Environ Pollut Barking Essex 1987 2014; 186: 1–6.
 Articles
30 Pope CA, Dockery DW. Air pollution and life expectancy in China
and beyond. Proc Natl Acad Sci USA 2013; 110: 12861–62.
31 Atkinson RW, Butland BK, Dimitroulopoulou C, et al.
Long-term exposure to ambient ozone and mortality: a quantitative
systematic review and meta-analysis of evidence from cohort
studies. BMJ Open 2016; 6: e009493.
32 Turner MC, Jerrett M, Pope CA, et al. Long-term ozone exposure
and mortality in a large prospective study. Am J Respir Crit Care Med
2016; 193: 1134–42.
33 Apte JS, Marshall JD, Cohen AJ, Brauer M. Addressing global
mortality from ambient PM2·5. Environ Sci Technol 2015; 49: 8057–66.
1918 www.thelancet.com Vol 389 May 13, 2017
34 Pope CA, Cropper M, Coggins J, Cohen A. Health benefits of air
pollution abatement policy: role of the shape of the
concentration–response function. J Air Waste Manag Assoc 1995
2015; 65: 516–22.
35 World Bank, Institute for Health Metrics and Evaluation. The cost
of air pollution: strengthening the economic case for action.
Washington, DC: World Bank, 2016. https://openknowledge.
worldbank.org/handle/10986/25013 (accessed March 3, 2017).
36 Correia AW, Pope CA, Dockery DW, Wang Y, Ezzati M, Dominici F.
Effect of air pollution control on life expectancy in the
United States. Epidemiology 2013; 24: 23–31.